{
  "context": {
    "cap_tokens": 129984,
    "max_model_len": 131072,
    "targets": [
      {
        "attempted_context_tokens": 129984,
        "clamped_tokens": 129984,
        "e2e_ms": 32770.31326293945,
        "estimated_prompt_tokens": 129984,
        "output_tokens": 64,
        "status": "passed",
        "target_tokens": 131072,
        "ttft_ms": 31905.324935913086,
        "usage": {
          "completion_tokens": 64,
          "prompt_tokens": 127426,
          "total_tokens": 127490
        }
      }
    ]
  },
  "failures": [],
  "identity": {
    "base_url": "http://127.0.0.1:39017/v1",
    "candidate_id": "qwen35-122b-a10b",
    "config_id": "vllm-nvfp4-131k",
    "model": "qwen35-122b-a10b-nvfp4",
    "started_at": "2026-07-11T00:26:01Z"
  },
  "intelligence": {
    "checks": [
      {
        "content_excerpt": "--- a/app.py\n+++ b/app.py\n@@ -1,3 +1,3 @@\n-timeout = 30\n+timeout = 45\n retries = 2",
        "id": "unified_diff_timeout_edit",
        "latency_ms": 603.0452251434326,
        "status": "passed",
        "text_checks": [
          {
            "name": "diff_shape",
            "passed": true
          },
          {
            "name": "removes_old_timeout",
            "passed": true
          },
          {
            "name": "adds_new_timeout",
            "passed": true
          }
        ],
        "validator": "deterministic_text_checks"
      },
      {
        "content_excerpt": "The total latency of 3000 ms exceeds the 2500 ms budget primarily due to the LLM's 1800 ms response time, which can be reduced by implementing streaming responses to overlap processing with audio gene",
        "id": "parallel_timeout_triage",
        "latency_ms": 752.957820892334,
        "status": "passed",
        "text_checks": [
          {
            "name": "identifies_budget_overrun",
            "passed": true
          },
          {
            "name": "offers_latency_fix",
            "passed": true
          }
        ],
        "validator": "deterministic_text_checks"
      }
    ],
    "status": "passed"
  },
  "run_id": "fast-bakeoff-20260711T002601Z",
  "schema": "anvil-serving.fast-tier-bakeoff/v1",
  "score_inputs": {
    "e2e_p50_ms": 32770.31326293945,
    "intelligence_pass_rate": 1.0,
    "operational_fit_notes": [
      "endpoint was already loaded; benchmark did not start or stop serves"
    ],
    "session_recall_passed": true,
    "thinking_mode": "disabled",
    "tool_call_passed": true,
    "ttft_p50_ms": 31905.324935913086,
    "usable_context_tokens": 131072,
    "voice_latency_ms": null
  },
  "selection": {
    "context_targets": [
      131072
    ],
    "endpoint_already_loaded": true,
    "requests_per_context": 1,
    "suites": [
      "chat",
      "context",
      "tool",
      "session",
      "intelligence"
    ]
  },
  "session": {
    "checks": [
      {
        "content_excerpt": "RIVER-918",
        "expected": "RIVER-918",
        "latency_ms": 168.95532608032227,
        "name": "single_request_multiturn_recall",
        "status": "passed"
      }
    ],
    "status": "passed"
  },
  "source_recipe": {
    "ref": "configs/serve-recipes.toml#nvidia-Qwen3.5-122B-A10B-NVFP4",
    "serve_command": "docker compose -f examples/primary-node/docker-compose.experiment.yml up -d heavy-qwen35-122b"
  },
  "thinking": {
    "chat_template_kwargs": {
      "enable_thinking": false
    },
    "mode": "disabled",
    "unsupported": false
  },
  "timing": {
    "chat": {
      "e2e_p50_ms": 32770.31326293945,
      "e2e_p95_ms": 32770.31326293945,
      "output_tokens": 64,
      "ttft_p50_ms": 31905.324935913086,
      "ttft_p95_ms": 31905.324935913086
    },
    "wall_ms": 34846.92120552063
  },
  "tool": {
    "checks": [
      {
        "arguments": {
          "zip": "98101"
        },
        "expected_arguments": {
          "zip": "98101"
        },
        "expected_function": "record_weather_zip",
        "latency_ms": 512.3310089111328,
        "name": "openai_tool_call_smoke",
        "status": "passed",
        "tool_call_count": 1,
        "valid_tool_call_count": 1,
        "validation_errors": []
      }
    ],
    "status": "passed"
  },
  "voice": {
    "llm_latency_ms": null,
    "status": "not_run",
    "stt_latency_ms": null,
    "total_turn_latency_ms": null,
    "tts_latency_ms": null
  }
}